U.S. patent number 10,554,560 [Application Number 15/711,417] was granted by the patent office on 2020-02-04 for predictive time allocation scheduling for computer networks.
This patent grant is currently assigned to Cisco Technology, Inc.. The grantee listed for this patent is Cisco Technology, Inc.. Invention is credited to Pascal Thubert, Jean-Philippe Vasseur, Patrick Wetterwald.
View All Diagrams
United States Patent |
10,554,560 |
Vasseur , et al. |
February 4, 2020 |
Predictive time allocation scheduling for computer networks
Abstract
In one embodiment, a device in a network receives data regarding
traffic volumes of deterministic and non-deterministic traffic
along a first path in the network. The device predicts, using the
received data, an increase in the traffic volume of the
non-deterministic traffic along the first path in the network. The
device identifies a period of time associated with the predicted
increase in the traffic volume of the non-deterministic traffic
along the first path. The device causes the deterministic traffic
to be sent along a second path in the network during the identified
period of time, to allow the first path to accommodate the
predicted increase in the traffic volume of the non-deterministic
traffic along the first path.
Inventors: |
Vasseur; Jean-Philippe (Saint
Martin d'uriage, FR), Thubert; Pascal (La Colle sur
Loup, FR), Wetterwald; Patrick (Mouans Sartoux,
FR) |
Applicant: |
Name |
City |
State |
Country |
Type |
Cisco Technology, Inc. |
San Jose |
CA |
US |
|
|
Assignee: |
Cisco Technology, Inc. (San
Jose, CA)
|
Family
ID: |
60989039 |
Appl.
No.: |
15/711,417 |
Filed: |
September 21, 2017 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20180026891 A1 |
Jan 25, 2018 |
|
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
|
14336250 |
Jul 21, 2014 |
9800506 |
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
H04L
47/25 (20130101); H04L 47/127 (20130101); H04L
45/22 (20130101); H04L 41/147 (20130101); H04L
47/17 (20130101); H04L 41/145 (20130101); H04L
47/823 (20130101) |
Current International
Class: |
H04L
12/801 (20130101); H04L 12/707 (20130101); H04L
12/911 (20130101); H04L 12/24 (20060101) |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
International Search Report dated Oct. 22, 2015 in connection with
PCT/US2015/040719. cited by applicant .
Dujovne et al., "6TiSCH On-the-Fly Scheduling
draft-dujovne-6tisch-on-the-fly-02", Feb. 14, 2014, pp. 1-10. cited
by applicant .
Thubert et al., "An Architecture for IPv6 over the TSCH mode of
IEEE 802.15.4e draft-ieft-6tisch-architecture-03", Jul. 4, 2014,
pp. 1-30. cited by applicant .
Vilajosana et al., "Minimal 6TiSCH Configuration
draft-ieft-6tisch-minimal-02", Jul. 4, 2014, pp. 1-20. cited by
applicant .
Palattella et al., "Terminology in IPv6 over the TSCH mode of IEEE
802.15.4e draft-ieft-6tisch-terminology-02", Jul. 4, 2014, pp.
1-12. cited by applicant .
Whatteyne et al., "Using IEEE802.15.4e TSCH in an LLN context:
Overview, Problem Statement and Goals draft-ieft-6tisch-tsch-01",
Jul. 4, 2014, pp. 1-22. cited by applicant.
|
Primary Examiner: Wong; Warner
Attorney, Agent or Firm: Behmke Innovation Group LLC Behmke;
James LeBarron; Stephen D.
Parent Case Text
RELATED APPLICATION
This application is a continuation-in-part of, and claims priority
to, U.S. patent application Ser. No. 14/336,250, filed on Jul. 21,
2014, entitled "Predictive Time Allocation Scheduling for TSCH
Networks," by Vasseur, et al., the contents of which are
incorporated herein by reference.
Claims
What is claimed is:
1. A method, comprising: receiving, at a device in a network, data
regarding traffic volumes of deterministic and non-deterministic
traffic along a first path in the network; predicting, by the
device and using the received data, an increase in the traffic
volume of the non-deterministic traffic along the first path in the
network; identifying, by the device, a period of time associated
with the predicted increase in the traffic volume of the
non-deterministic traffic along the first path; and causing, by the
device, the deterministic traffic to be sent along a second path in
the network during the identified period of time, to allow the
first path to accommodate the predicted increase in the traffic
volume of the non-deterministic traffic along the first path.
2. The method as in claim 1, further comprising: predicting, by the
device, a traffic volume of the deterministic traffic along the
path, based on the received data.
3. The method as in claim 1, wherein causing the deterministic
traffic to be sent along the second path in the network during the
identified period of time, to allow the first path to accommodate
the predicted increase in the traffic volume of the
non-deterministic traffic along the first path comprises:
unreserving, for the identified period of time, resources reserved
along the first path for the deterministic traffic.
4. The method as in claim 3, wherein the resources comprise one or
more communication time slots reserved for the deterministic
traffic along the first path or a bandwidth reserved for the
deterministic traffic.
5. The method as in claim 1, further comprising: causing, by the
device, the deterministic traffic to be sent via the first path
after the identified period of time.
6. The method as in claim 1, wherein the first path is a wired
network path.
7. The method as in claim 1, wherein predicting the increase in the
traffic volume of the non-deterministic traffic comprises: using a
machine learning-based model to model the traffic volume of the
non-deterministic traffic over time.
8. An apparatus, comprising: one or more network interfaces to
communicate with a network; a processor coupled to the network
interfaces and configured to execute one or more processes; and a
memory configured to store a process executable by the processor,
the process when executed configured to: receive data regarding
traffic volumes of deterministic and non-deterministic traffic
along a first path in the network; predict, using the received
data, an increase in the traffic volume of the non-deterministic
traffic along the first path in the network; identify a period of
time associated with the predicted increase in the traffic volume
of the non-deterministic traffic along the first path; and cause
the deterministic traffic to be sent along a second path in the
network during the identified period of time, to allow the first
path to accommodate the predicted increase in the traffic volume of
the non-deterministic traffic along the first path.
9. The apparatus as in claim 8, wherein the process when executed
is further configured to: predict a traffic volume of the
deterministic traffic along the path, based on the received
data.
10. The apparatus as in claim 8, wherein the device causes the
deterministic traffic to be sent along the second path in the
network during the identified period of time, to allow the first
path to accommodate the predicted increase in the traffic volume of
the non-deterministic traffic along the first path by: unreserving,
for the identified period of time, resources reserved along the
first path for the deterministic traffic.
11. The apparatus as in claim 10, wherein the resources comprise
one or more communication time slots reserved for the deterministic
traffic along the first path or a bandwidth reserved for the
deterministic traffic.
12. The apparatus as in claim 8, wherein the process when executed
further comprises: causing the deterministic traffic to be sent via
the first path after the identified period of time.
13. The apparatus as in claim 8, wherein the first path is a wired
network path.
14. The apparatus as in claim 8, wherein the apparatus predicts the
increase in the traffic volume of the non-deterministic traffic by:
using a machine learning-based model to model the traffic volume of
the non-deterministic traffic over time.
15. A tangible, non-transitory, computer-readable media having
software encoded thereon, wherein the software when executed by a
device in a network causes the device to perform a process
comprising: receiving, at the device in the network, data regarding
traffic volumes of deterministic and non-deterministic traffic
along a first path in the network; predicting, by the device and
using the received data, an increase in the traffic volume of the
non-deterministic traffic along the first path in the network;
identifying, by the device, a period of time associated with the
predicted increase in the traffic volume of the non-deterministic
traffic along the first path; and causing, by the device, the
deterministic traffic to be sent along a second path in the network
during the identified period of time, to allow the first path to
accommodate the predicted increase in the traffic volume of the
non-deterministic traffic along the first path.
16. The computer-readable media as in claim 15, wherein the process
further comprises: predicting, by the device, a traffic volume of
the deterministic traffic along the path, based on the received
data.
17. The computer-readable media as in claim 15, wherein causing the
deterministic traffic to be sent along the second path in the
network during the identified period of time, to allow the first
path to accommodate the predicted increase in the traffic volume of
the non-deterministic traffic along the first path comprises:
unreserving, for the identified period of time, resources reserved
along the first path for the deterministic traffic.
18. The computer-readable media as in claim 17, wherein the
resources comprise one or more communication time slots reserved
for the deterministic traffic along the first path or a bandwidth
reserved for the deterministic traffic.
19. The computer-readable media as in claim 15, the process further
comprising: causing, by the device, the deterministic traffic to be
sent via the first path after the identified period of time.
20. The computer-readable media as in claim 15, wherein the first
path is a wired network path.
Description
TECHNICAL FIELD
The present disclosure relates generally to computer networks, and,
more particularly, to predictive time allocation scheduling for
computer networks.
BACKGROUND
In general, deterministic networking attempts to precisely control
when a data packet arrives at its destination (e.g., within a
bounded timeframe). This category of networking may be used for a
myriad of applications such as industrial automation, vehicle
control systems, and other systems that require the precise
delivery of control commands to a controlled device. However,
implementing deterministic networking also places additional
requirements on a network. For example, packet delivery in a
deterministic network may require the network to exhibit fixed
latency, zero or near-zero jitter, and high packet delivery
ratios.
As an example of a deterministic network, consider a railway
system. A railway system can be seen as deterministic because
trains are scheduled to leave a railway station at certain times,
to traverse any number stations along a track at very precise
times, and to arrive at a destination station at an expected time.
From the human perspective, this is also done with virtually no
jitter. Which tracks are used by the different trains may also be
selected so as to prevent collisions and to avoid one train from
blocking the path of another train and delaying the blocked
train.
Low power and lossy networks (LLNs), e.g., Internet of Things (IoT)
networks, have a myriad of applications, such as sensor networks,
Smart Grids, and Smart Cities. Various challenges are presented
with LLNs, such as lossy links, low bandwidth, low quality
transceivers, battery operation, low memory and/or processing
capability, etc. Changing environmental conditions may also affect
device communications in an LLN. For example, physical obstructions
(e.g., changes in the foliage density of nearby trees, the opening
and closing of doors, etc.), changes in interference (e.g., from
other wireless networks or devices), propagation characteristics of
the media (e.g., temperature or humidity changes, etc.), and the
like, also present unique challenges to LLNs.
BRIEF DESCRIPTION OF THE DRAWINGS
The embodiments herein may be better understood by referring to the
following description in conjunction with the accompanying drawings
in which like reference numerals indicate identically or
functionally similar elements, of which:
FIG. 1 illustrates an example communication network;
FIG. 2 illustrates an example network device/node;
FIG. 3 illustrates an example message;
FIG. 4 illustrates an example directed acyclic graph (DAG) in the
communication network of FIG. 1;
FIG. 5 illustrates an example channel distribution/usage (CDU)
matrix;
FIG. 6 illustrates example chunks of the CDU matrix of FIG. 5;
FIGS. 7-8 illustrate examples of a parent node in the network of
FIG. 1 scheduling communications for a particular chunk;
FIGS. 9A-9C illustrate examples of time slot usages reports being
generated;
FIGS. 10A-E illustrate examples of time slot allocations being
adjusted based on usage predictions;
FIG. 11 illustrates an example simplified procedure for
predictively adjusting time slot assignment;
FIG. 12 illustrates an example simplified procedure for adjusting
time slot assignments of one or more child nodes;
FIG. 13 illustrates an example simplified procedure for generating
a time slot usage report; and
FIG. 14 illustrates an example simplified procedure for moving
seasonal deterministic traffic between network paths.
DESCRIPTION OF EXAMPLE EMBODIMENTS
Overview
According to one or more embodiments of the disclosure, a network
node provides one or more time slot usage reports to a time slot
usage prediction engine regarding a use of time slots of a channel
hopping schedule by one or more child nodes of the network node.
The network node receives a predicted time slot usage change for
the one or more child nodes. The network node generates one or more
updated time slot assignments for the one or more child nodes based
on the predicted time slot usage change. The network node provides
the one or more updated time slot assignments to the one or more
child nodes.
In further embodiments, a device in a network receives one or more
time slot usage reports regarding a use of time slots of a channel
hopping schedule by nodes in the network. The device predicts a
time slot demand change for a particular node based on the one or
more time slot usage reports. The device identifies a time frame
associated with the predicted time slot demand change. The device
adjusts a time slot assignment for the particular node in the
channel hopping schedule based on predicted demand change and the
identified time frame associated with the predicted time slot
demand change.
In another embodiment, a device in a network receives data
regarding traffic volumes of deterministic and non-deterministic
traffic along a first path in the network. The device predicts,
using the received data, an increase in the traffic volume of the
non-deterministic traffic along the first path in the network. The
device identifies a period of time associated with the predicted
increase in the traffic volume of the non-deterministic traffic
along the first path. The device causes the deterministic traffic
to be sent along a second path in the network during the identified
period of time, to allow the first path to accommodate the
predicted increase in the traffic volume of the non-deterministic
traffic along the first path.
Description
A computer network is a geographically distributed collection of
nodes interconnected by communication links and segments for
transporting data between end nodes, such as personal computers and
workstations, or other devices, such as sensors, etc. Many types of
networks are available, ranging from local area networks (LANs) to
wide area networks (WANs). LANs typically connect the nodes over
dedicated private communications links located in the same general
physical location, such as a building or campus. WANs, on the other
hand, typically connect geographically dispersed nodes over
long-distance communications links, such as common carrier
telephone lines, optical lightpaths, synchronous optical networks
(SONET), synchronous digital hierarchy (SDH) links, or Powerline
Communications (PLC) such as IEEE 61334, IEEE P1901.2, and others.
In addition, a Mobile Ad-Hoc Network (MANET) is a kind of wireless
ad-hoc network, which is generally considered a self-configuring
network of mobile routers (and associated hosts) connected by
wireless links, the union of which forms an arbitrary topology.
Smart object networks, such as sensor networks, in particular, are
a specific type of network having spatially distributed autonomous
devices such as sensors, actuators, etc., that cooperatively
monitor physical or environmental conditions at different
locations, such as, e.g., energy/power consumption, resource
consumption (e.g., water/gas/etc. for advanced metering
infrastructure or "AMI" applications) temperature, pressure,
vibration, sound, radiation, motion, pollutants, etc. Other types
of smart objects include actuators, e.g., responsible for turning
on/off an engine or perform any other actions. Sensor networks, a
type of smart object network, are typically shared-media networks,
such as wireless or PLC networks. That is, in addition to one or
more sensors, each sensor device (node) in a sensor network may
generally be equipped with a radio transceiver or other
communication port such as PLC, a microcontroller, and an energy
source, such as a battery. Often, smart object networks are
considered field area networks (FANs), neighborhood area networks
(NANs), etc. Generally, size and cost constraints on smart object
nodes (e.g., sensors) result in corresponding constraints on
resources such as energy, memory, computational speed and
bandwidth.
FIG. 1 is a schematic block diagram of an example computer network
100 illustratively comprising nodes/devices 200 (e.g., labeled as
shown, "FAR-1," `FAR-2," and "11," "12," . . . "46," and described
in FIG. 2 below) interconnected by various methods of
communication. For instance, the links 105 may be wired links or
shared media (e.g., wireless links, PLC links, etc.) where certain
nodes 200, such as, e.g., routers, sensors, computers, etc., may be
in communication with other nodes 200, e.g., based on distance,
signal strength, current operational status, location, etc. Those
skilled in the art will understand that any number of nodes,
devices, links, etc. may be used in the computer network, and that
the view shown herein is for simplicity. Also, those skilled in the
art will further understand that while network 100 is shown in a
certain orientation, particularly with a field area router (FAR)
node, the network 100 is merely an example illustration that is not
meant to limit the disclosure. Also as shown, a particular FAR
(e.g., FAR-1) may communicate via a WAN 130 with any number of
servers 150, such as a path computation element (PCE), network
management service (NMS), or other supervisory device.
Data packets 140 (e.g., traffic and/or messages sent between the
devices/nodes) may be exchanged among the nodes/devices of the
computer network 100 using predefined network communication
protocols such as certain known wired protocols, wireless protocols
(e.g., IEEE Std. 802.15.4, WiFi, Bluetooth.RTM., etc.), PLC
protocols, or other shared-media protocols where appropriate. In
this context, a protocol consists of a set of rules defining how
the nodes interact with each other. One communication technique
that may be used to implement links 105 is channel-hopping. Also
known as frequency hopping, use of such a technique generally
entails wireless devices "hopping" (e.g., alternating) between
different transmission and reception frequencies according to a
known schedule. Network 100 may also be divided into any number of
wireless domains (e.g., domains A-C) in which nodes 200 may
communicate.
FIG. 2 is a schematic block diagram of an example node/device 200
that may be used with one or more embodiments described herein,
e.g., as any of the nodes shown in FIG. 1 above. The device may
comprise one or more network interfaces 210 (e.g., wired, wireless,
PLC, etc.), at least one processor 220, and a memory 240
interconnected by a system bus 250, as well as a power supply 260
(e.g., battery, plug-in, etc.).
The network interface(s) 210, e.g., transceivers, include the
mechanical, electrical, and signaling circuitry for communicating
data over links 105 coupled to the network 100. The network
interfaces may be configured to transmit and/or receive data using
a variety of different communication protocols, particularly for
frequency-hopping communication as described herein. Note, further,
that the nodes may have two different types of network connections
210, e.g., wireless and wired/physical connections, and that the
view herein is merely for illustration. Also, while the network
interface 210 is shown separately from power supply 260, for PLC
the network interface 210 may communicate through the power supply
260, or may be an integral component of the power supply. In some
specific configurations the PLC signal may be coupled to the power
line feeding into the power supply.
The memory 240 includes a plurality of storage locations that are
addressable by the processor 220 and the network interfaces 210 for
storing software programs and data structures associated with the
embodiments described herein. Note that certain devices may have
limited memory or no memory (e.g., no memory for storage other than
for programs/processes operating on the device and associated
caches). The processor 220 may include hardware elements or
hardware logic configured to execute the software programs and
manipulate the data structures 245. An operating system 242,
portions of which are typically resident in memory 240 and executed
by the processor, functionally organizes the device by, inter alia,
invoking operations in support of software processes and/or
services executing on the device. These software processes and/or
services may include routing process/services 244, and an
illustrative channel hopping process 248 as described in greater
detail below. Note that while channel hopping process 248 is shown
in centralized memory 240, alternative embodiments provide for the
process to be specifically operated within the network interfaces
210, such as within a MAC layer 212 (as "process 248a").
It will be apparent to those skilled in the art that other
processor and memory types, including various computer-readable
media, may be used to store and execute program instructions
pertaining to the techniques described herein. Also, while the
description illustrates various processes, it is expressly
contemplated that various processes may be embodied as modules
configured to operate in accordance with the techniques herein
(e.g., according to the functionality of a similar process).
Further, while the processes have been shown separately, those
skilled in the art will appreciate that processes may be routines
or modules within other processes.
Routing process (services) 244 includes computer executable
instructions executed by the processor 220 to perform functions
provided by one or more routing protocols, such as proactive or
reactive routing protocols as will be understood by those skilled
in the art. These functions may, on capable devices, be configured
to manage a routing/forwarding table (a data structure 245)
including, e.g., data used to make routing/forwarding decisions. In
particular, in proactive routing, connectivity is discovered and
known prior to computing routes to any destination in the network,
e.g., link state routing such as Open Shortest Path First (OSPF),
or Intermediate-System-to-Intermediate-System (ISIS), or Optimized
Link State Routing (OLSR). Reactive routing, on the other hand,
discovers neighbors (i.e., does not have an a priori knowledge of
network topology), and in response to a needed route to a
destination, sends a route request into the network to determine
which neighboring node may be used to reach the desired
destination. Example reactive routing protocols may comprise Ad-hoc
On-demand Distance Vector (AODV), Dynamic Source Routing (DSR),
6LoWPAN Ad Hoc On-Demand Distance Vector Routing (LOAD), DYnamic
MANET On-demand Routing (DYMO), etc. Notably, on devices not
capable or configured to store routing entries, routing process 244
may consist solely of providing mechanisms necessary for source
routing techniques. That is, for source routing, other devices in
the network can tell the less capable devices exactly where to send
the packets, and the less capable devices simply forward the
packets as directed.
According to various embodiments, routing process 244 and/or
channel hopping process 248/248a may utilize machine learning
techniques, to predict a future state of the network (e.g., predict
routing changes, predict time slot usage by nodes, etc.). In
general, machine learning is concerned with the design and the
development of techniques that take as input empirical data (such
as network statistics and performance indicators), and recognize
complex patterns in these data. One very common pattern among
machine learning techniques is the use of an underlying model M,
whose parameters are optimized for minimizing the cost function
associated to M, given the input data. For instance, in the context
of classification, the model M may be a straight line that
separates the data into two classes such that M=a*x+b*y+c and the
cost function would be the number of misclassified points. The
learning process then operates by adjusting the parameters a,b,c
such that the number of misclassified points is minimal. After this
optimization phase (or learning phase), the model M can be used
very easily to classify new data points. Often, M is a statistical
model, and the cost function is inversely proportional to the
likelihood of M, given the input data.
As also noted above, learning machines (LMs) are computational
entities that rely one or more machine learning processes for
performing a task for which they haven't been explicitly programmed
to perform. In particular, LMs are capable of adjusting their
behavior to their environment. In the context of LLNs, and more
generally in the context of the IoT (or Internet of Everything,
IoE), this ability will be very important, as the network will face
changing conditions and requirements, and the network will become
too large for efficiently management by a network operator.
Artificial Neural Networks (ANNs) are a type of machine learning
technique whose underlying mathematical models that were developed
inspired by the hypothesis that mental activity consists primarily
of electrochemical activity between interconnected neurons. ANNs
are sets of computational units (neurons) connected by directed
weighted links. By combining the operations performed by neurons
and the weights applied by, ANNs are able to perform highly
non-linear operations to input data. The interesting aspect of
ANNs, though, is not that they can produce highly non-linear
outputs of the input, but that they can learn to reproduce a
predefined behavior through a training process. Accordingly, an ANN
may be trained to identify deviations in the behavior of a network
that could indicate the presence of a network attack (e.g., a
change in packet losses, link delays, number of requests,
etc.).
Low power and Lossy Networks (LLNs), e.g., certain sensor networks,
may be used in a myriad of applications such as for "Smart Grid"
and "Smart Cities." A number of challenges in LLNs have been
presented, such as:
1) Links are generally lossy, such that a Packet Delivery
Rate/Ratio (PDR) can dramatically vary due to various sources of
interferences, e.g., considerably affecting the bit error rate
(BER);
2) Links are generally low bandwidth, such that control plane
traffic must generally be bounded and negligible compared to the
low rate data traffic;
3) There are a number of use cases that require specifying a set of
link and node metrics, some of them being dynamic, thus requiring
specific smoothing functions to avoid routing instability,
considerably draining bandwidth and energy;
4) Constraint-routing may be required by some applications, e.g.,
to establish routing paths that will avoid non-encrypted links,
nodes running low on energy, etc.;
5) Scale of the networks may become very large, e.g., on the order
of several thousands to millions of nodes; and
6) Nodes may be constrained with a low memory, a reduced processing
capability, a low power supply (e.g., battery).
In other words, LLNs are a class of network in which both the
routers and their interconnect are constrained: LLN routers
typically operate with constraints, e.g., processing power, memory,
and/or energy (battery), and their interconnects are characterized
by, illustratively, high loss rates, low data rates, and/or
instability. LLNs are comprised of anything from a few dozen and up
to thousands or even millions of LLN routers, and support
point-to-point traffic (between devices inside the LLN),
point-to-multipoint traffic (from a central control point to a
subset of devices inside the LLN) and multipoint-to-point traffic
(from devices inside the LLN towards a central control point).
An example implementation of LLNs is an "Internet of Things"
network. Loosely, the term "Internet of Things" or "IoT" may be
used by those in the art to refer to uniquely identifiable objects
(things) and their virtual representations in a network-based
architecture. In particular, the next frontier in the evolution of
the Internet is the ability to connect more than just computers and
communications devices, but rather the ability to connect "objects"
in general, such as lights, appliances, vehicles, HVAC (heating,
ventilating, and air-conditioning), windows and window shades and
blinds, doors, locks, etc. The "Internet of Things" thus generally
refers to the interconnection of objects (e.g., smart objects),
such as sensors and actuators, over a computer network (e.g., IP),
which may be the Public Internet or a private network. Such devices
have been used in the industry for decades, usually in the form of
non-IP or proprietary protocols that are connected to IP networks
by way of protocol translation gateways. With the emergence of a
myriad of applications, such as the smart grid, smart cities, and
building and industrial automation, and cars (e.g., that can
interconnect millions of objects for sensing things like power
quality, tire pressure, and temperature and that can actuate
engines and lights), it has been of the utmost importance to extend
the IP protocol suite for these networks.
An example protocol specified in an Internet Engineering Task Force
(IETF) Proposed Standard, Request for Comment (RFC) 6550, entitled
"RPL: IPv6 Routing Protocol for Low Power and Lossy Networks" by
Winter, et al. (March 2012), provides a mechanism that supports
multipoint-to-point (MP2P) traffic from devices inside the LLN
towards a central control point (e.g., LLN Border Routers (LBRs) or
"root nodes/devices" generally), as well as point-to-multipoint
(P2MP) traffic from the central control point to the devices inside
the LLN (and also point-to-point, or "P2P" traffic). RPL
(pronounced "ripple") may generally be described as a distance
vector routing protocol that builds a Directed Acyclic Graph (DAG)
for use in routing traffic/packets 140, in addition to defining a
set of features to bound the control traffic, support repair, etc.
Notably, as may be appreciated by those skilled in the art, RPL
also supports the concept of Multi-Topology-Routing (MTR), whereby
multiple DAGs can be built to carry traffic according to individual
requirements.
A DAG is a directed graph having the property that all edges
(and/or vertices) are oriented in such a way that no cycles (loops)
are supposed to exist. All edges are included in paths oriented
toward and terminating at one or more root nodes (e.g.,
"clusterheads or "sinks"), often to interconnect the devices of the
DAG with a larger infrastructure, such as the Internet, a wide area
network, or other domain. In addition, a Destination Oriented DAG
(DODAG) is a DAG rooted at a single destination, i.e., at a single
DAG root with no outgoing edges. A "parent" of a particular node
within a DAG is an immediate successor of the particular node on a
path towards the DAG root, such that the parent has a lower "rank"
than the particular node itself, where the rank of a node
identifies the node's position with respect to a DAG root (e.g.,
the farther away a node is from a root, the higher is the rank of
that node). Further, in certain embodiments, a sibling of a node
within a DAG may be defined as any neighboring node which is
located at the same rank within a DAG. Note that siblings do not
necessarily share a common parent, and routes between siblings are
generally not part of a DAG since there is no forward progress
(their rank is the same). Note also that a tree is a kind of DAG,
where each device/node in the DAG generally has one parent or one
preferred parent.
DAGs may generally be built (e.g., by routing process 244) based on
an Objective Function (OF). The role of the Objective Function is
generally to specify rules on how to build the DAG (e.g. number of
parents, backup parents, etc.).
In addition, one or more metrics/constraints may be advertised by
the routing protocol to optimize the DAG against. Also, the routing
protocol allows for including an optional set of constraints to
compute a constrained path, such as if a link or a node does not
satisfy a required constraint, it is "pruned" from the candidate
list when computing the best path. (Alternatively, the constraints
and metrics may be separated from the OF.) Additionally, the
routing protocol may include a "goal" that defines a host or set of
hosts, such as a host serving as a data collection point, or a
gateway providing connectivity to an external infrastructure, where
a DAG's primary objective is to have the devices within the DAG be
able to reach the goal. In the case where a node is unable to
comply with an objective function or does not understand or support
the advertised metric, it may be configured to join a DAG as a leaf
node. As used herein, the various metrics, constraints, policies,
etc., are considered "DAG parameters."
Illustratively, example metrics used to select paths (e.g.,
preferred parents) may comprise cost, delay, latency, bandwidth,
expected transmission count (ETX), etc., while example constraints
that may be placed on the route selection may comprise various
reliability thresholds, restrictions on battery operation,
multipath diversity, bandwidth requirements, transmission types
(e.g., wired, wireless, etc.). The OF may provide rules defining
the load balancing requirements, such as a number of selected
parents (e.g., single parent trees or multi-parent DAGs). Notably,
an example for how routing metrics and constraints may be obtained
may be found in an IETF RFC, entitled "Routing Metrics used for
Path Calculation in Low Power and Lossy Networks" <RFC 6551>
by Vasseur, et al. (March 2012). Further, an example OF (e.g., a
default OF) may be found in an IETF RFC, entitled "RPL Objective
Function 0" <RFC 6552> by Thubert (March 2012) and "The
Minimum Rank Objective Function with Hysteresis" <RFC 6719>
by O. Gnawali et al. (September 2012).
Building a DAG may utilize a discovery mechanism to build a logical
representation of the network, and route dissemination to establish
state within the network so that routers know how to forward
packets toward their ultimate destination. Note that a "router"
refers to a device that can forward as well as generate traffic,
while a "host" refers to a device that can generate but does not
forward traffic. Also, a "leaf" may be used to generally describe a
non-router that is connected to a DAG by one or more routers, but
cannot itself forward traffic received on the DAG to another router
on the DAG. Control messages may be transmitted among the devices
within the network for discovery and route dissemination when
building a DAG.
According to the illustrative RPL protocol, a DODAG Information
Object (DIO) is a type of DAG discovery message that carries
information that allows a node to discover a RPL Instance, learn
its configuration parameters, select a DODAG parent set, and
maintain the upward routing topology. In addition, a Destination
Advertisement Object (DAO) is a type of DAG discovery reply message
that conveys destination information upwards along the DODAG so
that a DODAG root (and other intermediate nodes) can provision
downward routes. A DAO message includes prefix information to
identify destinations, a capability to record routes in support of
source routing, and information to determine the freshness of a
particular advertisement. Notably, "upward" or "up" paths are
routes that lead in the direction from leaf nodes towards DAG
roots, e.g., following the orientation of the edges within the DAG.
Conversely, "downward" or "down" paths are routes that lead in the
direction from DAG roots towards leaf nodes, e.g., generally going
in the opposite direction to the upward messages within the
DAG.
Generally, a DAG discovery request (e.g., DIO) message is
transmitted from the root device(s) of the DAG downward toward the
leaves, informing each successive receiving device how to reach the
root device (that is, from where the request is received is
generally the direction of the root). Accordingly, a DAG is created
in the upward direction toward the root device. The DAG discovery
reply (e.g., DAO) may then be returned from the leaves to the root
device(s) (unless unnecessary, such as for UP flows only),
informing each successive receiving device in the other direction
how to reach the leaves for downward routes. Nodes that are capable
of maintaining routing state may aggregate routes from DAO messages
that they receive before transmitting a DAO message. Nodes that are
not capable of maintaining routing state, however, may attach a
next-hop parent address. The DAO message is then sent directly to
the DODAG root that can in turn build the topology and locally
compute downward routes to all nodes in the DODAG. Such nodes are
then reachable using source routing techniques over regions of the
DAG that are incapable of storing downward routing state. In
addition, RPL also specifies a message called the DIS (DODAG
Information Solicitation) message that is sent under specific
circumstances so as to discover DAG neighbors and join a DAG or
restore connectivity.
FIG. 3 illustrates an example simplified control message format 300
that may be used for discovery and route dissemination when
building a DAG, e.g., as a DIO, DAO, or DIS message. Message 300
illustratively comprises a header 310 with one or more fields 312
that identify the type of message (e.g., a RPL control message),
and a specific code indicating the specific type of message, e.g.,
a DIO, DAO, or DIS. Within the body/payload 320 of the message may
be a plurality of fields used to relay the pertinent information.
In particular, the fields may comprise various flags/bits 321, a
sequence number 322, a rank value 323, an instance ID 324, a DODAG
ID 325, and other fields, each as may be appreciated in more detail
by those skilled in the art. Further, for DAO messages, additional
fields for destination prefixes 326 and a transit information field
327 may also be included, among others (e.g., DAO_Sequence used for
ACKs, etc.). For any type of message 300, one or more additional
sub-option fields 328 may be used to supply additional or custom
information within the message 300. For instance, an objective code
point (OCP) sub-option field may be used within a DIO to carry
codes specifying a particular objective function (OF) to be used
for building the associated DAG. Alternatively, sub-option fields
328 may be used to carry other certain information within a message
300, such as indications, requests, capabilities, lists,
notifications, etc., as may be described herein, e.g., in one or
more type-length-value (TLV) fields.
FIG. 4 illustrates an example simplified DAG 400 that may be
created, e.g., through the techniques described above, within
network 100 of FIG. 1. For instance, certain links 105 may be
selected for each node to communicate with a particular parent (and
thus, in the reverse, to communicate with a child, if one exists).
These selected links form the DAG 400 (shown as bolded lines),
which extends from the root node toward one or more leaf nodes
(nodes without children). Traffic/packets 140 (shown in FIG. 1) may
then traverse the DAG 400 in either the upward direction toward the
root or downward toward the leaf nodes, particularly as described
herein.
According to various embodiments, communications within network 100
may be deterministic. Notably, low power wireless industrial
process control typically uses 1 Hz to 4 Hz control loops, and for
those, a scheduled MAC protocol can be considered deterministic,
even when clocks drift in the order of tens of parts per million
(ppm). A low-throughput technology such as IEEE 802.15.4 may thus
be adapted to support determinism. In particular, the bandwidth can
be pre-formatted in a time division multiplexing (TDM) fashion
using IEEE 802.15.4, and time slots become a unit of throughput
that can allocated to a deterministic flow, without incurring a
huge consumption of system resources. In other implementations of a
time sensitive network, individual timers may be used by the
networked devices instead of TDM. Such timers may elapse at the
time of a deterministic transmission, so as to reserve the medium
for that transmission, leaving the medium free for best effort
routing the rest of the time.
Routing in a deterministic network can be operated either in a
centralized or in a distributed fashion, but only the centralized
routing operation can guarantee the overall optimization for all
the flows with a given set of constraints and goals. An example
architecture to support such a technique may be found in the IETF
draft entitled "An Architecture for IPv6 over the TSCH mode of IEEE
802.15.4e" by Thubert et al. (February 2014), and referred to
herein as "6TiSCH". The centralized computation is typically done
by a PCE with an objective function that represents the goals and
constraints. A PCE may compute not only an optimized Layer 3 path
for purposes of traffic engineering, but also to compute time slots
associated with a deterministic flow at the same time as it
computes a route over an LLN. Generally speaking, this requires the
PCE to have knowledge of the flows as well as knowledge of the
radio behavior at each hop (e.g., an estimation of the expected
transmission count (ETX) so as to provision enough time slots for
retransmissions).
For distributed routing, 6TiSCH relies on the RPL routing protocol
(RFC6550). The design of RPL also includes the capability to build
routing topologies (e.g., "instances" in RPL parlance) that are
associated with objective functions, but in a distributed fashion.
With RPL, the routing operations will be more efficient (e.g., with
no need of CPU intensive PCE computations) and resilient (e.g.,
with no dependence on a PCE for base routing and recovery).
Of note is that scheduling is not a part of RPL and may be designed
for the distributed routing scheme. Although it is not possible to
guarantee that an individual path is fully optimized, or that the
distribution of resources is globally optimized, it may be possible
to impose deterministic behavior along a routing path (e.g., an
ultra-low jitter, controlled latency, etc.).
For the underlying MAC operation, 6TiSCH relies, as its name shows,
on time slotted channel hopping (TSCH). More specifically, 6TiSCH
is being designed for the IEEE 802.15.4e TSCH mode of operation.
This is the standardized version of the MAC that was adopted by all
industrial WSN standards, ISA 100.11a, WirelessHART and WIAPA.
The time slotted aspect of the TSCH technology is a time division
multiplexing (TDM) technique, which requires all nodes in the
network to be time synchronized. In other words, time is sliced up
into time slots with a given time slot being long enough for a MAC
frame of maximum size to be sent from mote B to node A, and for
node A to reply with an acknowledgment (ACK) frame indicating
successful reception.
TSCH is different from traditional low-power MAC protocols because
of its scheduled nature. In TSCH, all nodes in the network follow a
common communication schedule, which indicates for each active
(e.g., transmit or receive) timeslot a channel offset and the
address of the neighbor to communicate with. The channel offset is
translated into a frequency using a specific translation function
which causes pairs of neighbors to "hop" between the different
available channels (e.g., frequencies) when communicating. Such
channel hopping technique efficiently combats multi-path fading and
external interference. Notably, since 6TiSCH is based on TSCH,
6TiSCH also uses the basic TSCH concepts of a schedule and time
slots. However, since 6TiSCH may extend over several interference
domains with distributed routing and scheduling, there is no longer
the concept of a single schedule that would centralize all the
transmissions and receptions. In particular, with 6TiSCH, some TSCH
concepts may still apply globally and their configurations must be
shared by all nodes in the network, but other concepts may be local
to a given node in 6TiSCH. For example, the hopping schedule in
6TiSCH may represent only the transmission to which a particular
node is participating.
Referring now to FIG. 5, an example channel distribution/usage
(CDU) matrix 500 is shown that may be used by the nodes/devices 200
in network 100. Notably, 6TiSCH defines a new global concept of a
CDU matrix that may repeat itself over time and represents the
global characteristics of the network such as used/unused channels,
timeslot durations, number of time slots per iteration, etc. As
shown, CDU matrix 500 may include an index of channel offsets 502
along a first axis that correspond to the channels available for
use in network 100 (e.g., offsets for each of sixteen available
channels). As would be appreciated, any number of channels may be
used in the network. Along the other axis are slot offsets 504 that
correspond to differing time slots, the combination of which is
equal to one period of the network scheduling operation.
CDU matrix 500 may be used to define the basic wireless
communication operations for the network. For example, CDU matrix
500 may be used to define the duration of a timeslot (e.g., between
10 to 15 ms), the period of an iteration (e.g., the total number of
time slots, indexed by slot offsets 504), and the number of
channels (e.g., indexed by channel offset 502) to which the MAC may
jump.
A "cell" in CDU matrix 500 is defined by the pair (slot offset,
channel offset) in the epochal description of CDU matrix 500, in
other words, at time t=0. During runtime, the actual channel at
which a given transmission happens may be rotated to avoid
interferences such as self-inflicted multipath fading.
Referring now to FIG. 6, an example subset 600 of CDU matrix 500 is
shown to be divided into chunks 606. In order to scale the network,
the computation of the channel hopping schedule for the network may
be distributed. According to some embodiments, such as those in
which 6TiSCH is used, a parent node (e.g., an RPL parent) may be
responsible for computing the schedule between the parent and its
child node(s) in both directions. In order to allocate a cell for a
given transmission, the parent node must be certain that this cell
will not be used by another parent in the interference domain. As
shown, for example, cells within CDU matrix 500 may be "owned" by
different parent nodes within the network. The collective cells of
CDU matrix 500 assigned to different parent nodes may then be
grouped together as chunks 606. In one implementation, for example,
CDU matrix 500 may be formatted into chunks by making a full
partition of matrix 500. The resulting partition must be well known
by all the nodes in the network, to support the appropriation
process, which would rely on a negotiation between nodes within an
interference domain.
Typically, there will be at most one cell in a chunk per column of
CDU matrix 500, to reflect that a device with a single radio may
not use two channels at the same time. The cells may also be well
distributed in time and frequency, so as to limit the gaps between
transmissions and avoid the sequential loss of frames in multipath
fading due to the consecutive reuse of a same channel.
Chunks 606 may be defined at the epochal time (e.g., at the time of
creation of CDU matrix 500) and the 802.15.4e operation may be
repeated iteratively any number of times. Typically, the effective
channel for a given transmission may be incremented by a constant
that is prime with the number of channels, modulo the number of
channels at each iteration. As a result, the channel of a given
transmission changes at each iteration and the matrix virtually
rotates.
FIGS. 7-8 illustrate examples of a parent node in the network of
FIG. 1 scheduling communications for a particular chunk. As shown,
assume that node 32 is the parent node of child nodes 41, 42
according to the routing protocol. Node 32 may be assigned a chunk
(e.g., chunk A) of CDU matrix 500, thereby allowing node 32 to
manage the usage of the corresponding cells in the chunk within its
interference domain. Thus, node 32 may decide which transmissions
will occur over the cells in the chunk between itself and its child
node(s). Ultimately, a chunk represents some amount of bandwidth
and can be seen as the generalization in the time/frequency domain
of the classical channel that is used to paint a wireless
connectivity graph, e.g. to distribute TV frequencies over a
country or WiFi channels in an ESS deployment.
If chunks are designed to form a partition of the CDU matrix 500,
multiple different chunks may be in use in the same area of network
100 and under the control of different parents. In one embodiment,
the appropriation process may be such that any given node that
communicates using cells in a given chunk, as appropriated and
managed by a parent A, should not be within the interference domain
of any other node that is also communicating using the same chunk
but appropriated and managed by a different parent B. Consequently,
the number of parents in any given area of the network may be
constrained by the number of chunks.
Referring more specifically to FIG. 8, parent node 32 may use a
slot frame 802 to assign hopping schedules 804, 806 to itself and
any of its child node(s), respectively. Generally speaking, slot
frame 802 is a MAC-level abstraction that is also internal to the
node and includes a series of time slots of equal length and
priority. For example, the size of the slot frame 802 may match the
CDU matrix 500. Parent node 32 may use slot frame 802 to divide the
corresponding times into slots and associate the slots to a
particular operation (e.g., reception, transmission, multicast
operation, etc.). For example, as shown, parent node 32 and one of
its child nodes may be synchronized to use the same channel during
a given time slot.
Slot frame 802 may be characterized by a slotframe_ID, a slot
duration, and a slotframe_size. In some implementations, multiple
slot frames may coexist in a node's schedule. In other words, a
node can have multiple activities scheduled in different slot
frames, based on the priority of its packets/traffic flows. The
different slot frames may be implemented as having the same
durations/sizes or different durations/sizes, in various cases. The
time slots in the slot frame may also be indexed by the slot
offsets 604 (e.g., the first time slot in slot frame 802 may be
indexed as slot offset 0, etc.).
In various implementations, different parent nodes may appropriate
different chunks such that the chunks used throughout the network
do not interfere. For example, chunks may be appropriated by
different parent nodes such that, for a given chunk, the domains do
not intersect. In addition, the domains for different chunks are
generally not congruent since the chunks are owned by different
nodes. As a result, the schedule in a node with a single radio is a
series of transmissions, and the parent to child cells are taken
from (one of) the chunk(s) that the parent has appropriated.
6TiSCH also defines the peer-wise concept of a "bundle," that is
needed for the communication between adjacent nodes. In general, a
bundle is a group of equivalent scheduled cells (e.g., cells
identified by different slot offset/channel offset pairs), which
are scheduled for a same purpose, with the same neighbor, with the
same flags, and the same slot frame. The size of the bundle refers
to the number of cells it includes. Given the length of the slot
frame, the size of the bundle also translates directly into
bandwidth, either logical or physical. Ultimately a bundle
represents a half-duplex link between nodes, one transmitter and
one or more receivers, with a bandwidth that amount to the sum of
the time slots in the bundle. Adding a timeslot in a bundle
increases the bandwidth of the link.
Track forwarding is the simplest and fastest forwarding model
defined in the 6TiSCH architecture that specifies IPv6 over TSCH.
In general, a "track" is defined as an end-to-end succession of
time slots, with a particular timeslot belonging to at most one
track. In this model, a set of input cells (time slots) are
uniquely bound to a set of output cells, representing a forwarding
state that can be used regardless of the upper layer protocol. This
model can effectively be seen as a G-MPLS operation in that the
information used to switch is not an explicit label, but rather
related to other properties of the way the packet was received, a
particular cell in the case of 6TiSCH. As a result, as long as the
TSCH MAC (and Layer 2 security) accepts a frame, that frame can be
switched regardless of the protocol, whether this is an IPv6
packet, a 6LoWPAN fragment, or a frame from an alternate protocol
such as WirelessHART of ISA 100.11a.
For a given iteration of a slotframe, the timeslot is associated
uniquely with a cell, which indicates the channel at which the
timeslot operates for that iteration. A data frame that is
forwarded along a track has a destination MAC address set to
broadcast or a multicast address depending on MAC support. This
way, the MAC layer in the intermediate nodes accepts the incoming
frame and the 6 top sublayer switches it without incurring a change
in the MAC header. In the case of IEEE 802.15.4e, this means
effectively broadcast, so that along the Track the short address
for the destination is set to broadcast, 0xFFFF. Conversely, a
frame that is received along a track with a destination MAC address
set to this node is extracted from the track stream and delivered
to the upper layer. A frame with an unrecognized MAC address may be
ignored at the MAC layer and thus is not received at the 6 top
sublayer.
As noted above, scheduling communications in a deterministic
network may be difficult, particularly when the network is
scalable. Notably, centralized computation of time schedules by a
network device (e.g., a PCE, etc.) may require a priori knowledge
of the traffic demands between all nodes in the networks. Although
on-the-fly real time traffic reporting may be implemented in a
network, the additional bandwidth requirements to do so may be
unsuitable for many situations. For example, while typically not a
fundamental issue in high-bandwidth network such as IP/MPLS
networks, sending such communications to a PCE may present a
significant issue for scaling a TSCH network. In particular, if the
presence of a super-flow is detected in the network, a node may
trigger a request to the PCE. In response, the PCE may re-compute
time schedules for a potentially large number of existing flows
(which is known as an NP-Complete problem), before sending back the
new schedules. In a constrained network such as an LLN, the
resulting control plane and response time may be unacceptable for
many applications. Additionally, if certain nodes are battery
operated, the additional traffic associated with the real time
reporting may directly impinge the life expectancy of the network
devices.
Predictive Time Allocation Scheduling for Computer Networks
The techniques herein provide a machine learning-based architecture
that may make time slot allocation changes based on predicted
traffic changes. In some aspects, information regarding actual
traffic and time slot usage by the network nodes may be used to
adjust time slot allocations (e.g., allocated cells of a CDU
matrix) in relation to a predicted burst of traffic. The machine
learning model may, in some cases, be hosted on a centralized
network device (e.g., a PCE, etc.) and receive time slot usage
reports on a per-child-basis, along with other network statistics,
such as the queuing delays. In one embodiment, predictions of
traffic changes and/or their seasonality may be made using
background processing instead of being explicitly requested by
nodes in the network. In turn, the PCE may trigger the dynamic
allocation or removal of time slots from the nodes or provide in
advance the computed time frames according to its own
prediction.
Specifically, according to one or more embodiments of the
disclosure as described in detail below, a device in a network
receives one or more time slot usage reports regarding a use of
time slots of a channel hopping schedule by nodes in the network.
The device predicts a time slot demand change for a particular node
based on the one or more time slot usage reports. The device
identifies a time frame associated with the predicted time slot
demand change. The device adjusts a time slot assignment for the
particular node in the channel hopping schedule based on predicted
demand change and the identified time frame associated with the
predicted time slot demand change.
Illustratively, the techniques described herein may be performed by
hardware, software, and/or firmware, such as in accordance with the
channel hopping process 248/248a, which may include computer
executable instructions executed by the processor 220 (or
independent processor of interfaces 210) to perform functions
relating to the techniques described herein, e.g., in conjunction
with routing process 244. For example, the techniques herein may be
treated as extensions to conventional protocols, such as the
various PLC protocols or wireless communication protocols (e.g.,
IEEE 802.15.4e 6TiSCH, etc.), and as such, may be processed by
similar components understood in the art that execute those
protocols, accordingly.
Operationally, a predictive approach may be taken by a centralized
networking device (e.g., a PCE, etc.), to perform time scheduling
that takes into account predicted traffic changes and/or any
associated seasonality of the traffic changes. As used herein, the
centralized device may be referred to as a Predictive Time
Scheduler (PTS) PCE (e.g., a prediction engine). In contrast to a
PCE that allocates time slots according to a priori knowledge of
the traffic flows or in response to an explicit request (e.g., from
the nodes, from an NMS, etc.), a PTS may make base time slot
allocations on network conditions predicted by a machine learning
model. Example models may include, but are not limited to,
auto-regressive moving average (ARMA) models, ARMA-X models that
take into account exogenous variables (X), Hidden Markov Models
(HMMs), Gaussian Processes, or any other machine learning model
that can be used to predict traffic demand changes and/or the
seasonality of such changes. Although described primarily using the
example of adding more time slots to a node, the techniques herein
may be applied in a similar manner to remove time slots from a
node, if its traffic is predicted to decrease.
In some embodiments, each node may be configured to provide
compressed information regarding use of its allocated time slots
within a time slot usage report. As noted previously, each parent
node may receive or send packets between itself and a child node
according to a given routing topology (e.g., a DODAG computed by a
distributed routing protocol such as RPL), with one packet and
acknowledgement per time slot.
Referring now to FIGS. 9A-9C, examples are shown of time slot
usages reports being generated. For example, as shown in FIG. 9A,
parent node 32 may monitor the use of the time slots allocated to
its child nodes 41 and 42. Based on the traffic send during the
allocated time slots between parent node 32 and child nodes 41 and
42, parent node 32 may then generate a time slot usage report that
quantifies how heavily the nodes use the time slots.
In some cases, a time slot usage report may also include
information regarding queuing delays experienced by a child node
and/or alternative paths used by the child node (e.g., due to a
time slot with its preferred parent being unavailable). For
example, as shown in FIG. 9B, child node 42 may send a message 902
to its parent node 32 that provides feedback regarding any queuing
delays experienced by node 42. In particular, if node 42 is running
out of allocated time slots to send traffic to its parent node 32,
node 42 may queue the traffic until its next available time slot,
thereby delaying the traffic. Since node 32 is otherwise unaware of
the queuing delay experienced by the user traffic from its children
nodes 41 and 42, notification message 902 may be sent to parent
node 32, to report on any experienced queuing delays. In some
cases, a reported delay may also include information regarding a
priority associated with the delayed traffic.
In some embodiments, message 902 may be a custom IPv6 link local
message or a custom type-length-value (TLV) piggybacked using the
routing protocol. For example, if the routing protocol in use is
RPL in storing mode, message 902 may be a DAO message that includes
queuing delay information within a TLV, which may be consumed by
the route storing parent node 32. In another embodiment, message
902 may be piggybacked with a data frame as an IEEE 802.15.4e
Information Element (IE). In response to receiving message 902, the
parent node 32 may include any delay-related information from
message 902 in a time slot usage report. Notably, such information
may be used by the prediction engine in its predictions. For
example, if the machine learning model used by the prediction
engine detects increased delays from node 42 to its parent node 32,
this may be used by the prediction engine to detect an increase in
traffic. Detecting a time-based pattern in increased delays may
also be treated by the prediction engine as a sign of seasonality
and used by the prediction engine to allocate more time slots for
the affected nodes at the specific period of time.
As shown in FIG. 9C, node 32 may provide a time slot usage report
904 to the prediction engine (e.g., a PCE 150). In one embodiment,
each node sends after the expiration of a timer T, a time slot
usage report to the PTS that contains a bit map of the set of time
slots that were effectively used by each of its children. In
another embodiment, the report can be sent if the node determines
that the proportion of time slots effectively being used has
changed significantly. In yet another embodiment, the periodic
timer T may be dynamically computed by the PTS according to the
prediction accuracy of the computed predictive model. For example,
if the PTS determines that the time slot usage matches its
prediction, it may increase the periodic timer T, to extend the
periodicity of the usage reports. Conversely, the timer T may be
reduced to increase the frequency of the reports, should the
resulting predictions prove to be inaccurate.
FIGS. 10A-E illustrate examples of time slot allocations being
adjusted based on usage predictions, according to various
embodiments. In response to receiving the time slot usage reports,
the prediction engine (e.g., a PCE in servers 150) may then perform
predictions on the slot usage, as shown in FIG. 10A. In some
implementations, the prediction engine may perform off-line
background optimization of time slots in light of a prediction of
all traffic between each node and their parent within a given
collision domain. For example, if the prediction engine determines
that in X hours the number of slots between node 42 and its parent
node 32 will increase by x % while the traffic between node 41 and
its parent node 32 will decrease by y %, the prediction engine may
perform time slot arbitration accordingly (e.g., by reallocating
some time slots from node 41 to node 42). Even in the absence of
arbitration, if the prediction engine determines that the number of
time slots needed between a pair of nodes is likely to increase
over time or may be increased for a specific period of time (e.g.,
for two hours, on each Friday between 4 PM and 5 PM, etc.), the
prediction engine may thus be able to anticipate time-based
scheduling.
In all cases, two different modes of operation may be used to
adjust the time schedules of the network nodes. In a first mode of
operation, the PTS explicitly allocates new time slots to the
various nodes in the network according to its prediction and using
the existing time schedule-based approach. In other words, the time
schedules are uploaded similarly to the existing approach but
instead of reacting to explicit request, it is the PTS/prediction
engine that makes use of an unsolicited action of time slot
provisioning according to its predictions. For example, as shown in
FIG. 10B, the prediction engine may explicitly allocate new time
slots to the various nodes via an instruction 1002.
In another mode of operation, individual nodes may plan the
communication schedule changes. In this mode, a node may be allowed
to push a set of time-based schedules according to the predictions
made by the prediction engine. For example, as shown in FIG. 10C,
the PTS may send a message 1004 to node 32. Message 1004 may
indicate to node 32 the state of a time frame and its upcoming
changes at times T1, T2, etc., without the prediction engine having
to resend the schedule. If the PTS predicts a traffic burst at T2-x
milliseconds, it would then send the time frame while mentioning
the values of the time frame at T2, with x being the time to
allocate more time slots between the node and its parents. In
response, node 32 may send instructions 1006 to its child nodes, to
update their time slot allocations. Thus, the time slot assignment
updates may be implemented prior to the predicted burst of
traffic.
In some cases, a closed-loop mechanism may be implemented between
the prediction engine and the nodes, as shown in FIG. 10E. In
particular, the prediction engine may receive usage reports 904
from the nodes after a predicted change, to determine how accurate
the prediction was. Indeed, if the prediction engine increases or
decreases the time slot allocations between a pair of nodes based
on a predicted traffic change, the prediction engine may monitor
how accurate the prediction was and send further allocation changes
as needed. For example, the prediction engine may adjust the
periodic timer T used by the affected nodes, to increase the
frequency of the time slot usage reports sent by the nodes to the
prediction engine.
In various embodiments, the techniques herein can also be extended
to other forms of networks, such as wired networks, non-TSCH
wireless networks, and the like. In particular, deterministic
networking generally relies on a reservation mechanism that
allocates physical resources to particular constant bit rate (CBR)
flows at particular times and along a particular network path.
However, not all flows are CBR flows, thus wasting bandwidth.
In addition, as noted above, any additional resources along a
network path that are not reserved for deterministic traffic may be
utilized for non-deterministic traffic that may be sent in a best
effort manner, instead. Note that some best effort traffic may
still be quite critical, even if it does not fit the requirements
for deterministic resource reservations (e.g., the traffic is
highly variable, not time sensitive, etc.). Indeed, the amount of
bandwidth needed by best effort traffic can be variable, far more
so than deterministic traffic. Thus, the location in the network at
which the peak amount of resources is consumed by best effort
traffic can also change over time. However, a heavy use of
deterministic reservations may leave little to no additional
resources available along a path for best effort traffic.
As described above, machine learning can be utilized to assess the
seasonality of the deterministic and/or non-deterministic (e.g.,
best effort) traffic flows in a network. In further embodiments,
this information can then be used by the system to identify traffic
flows that do not exhibit any seasonality or are only weakly
seasonal (e.g., below a seasonality threshold). In turn, the system
can reroute the flows without seasonality in the long term and/or
reroute the seasonal traffic for short amounts of time, to free up
resources for peak utilization by best effort traffic.
In a further embodiment, the system may compute alternate
deterministic paths, particularly for flows that exhibit
predictable seasonal behaviours. In other words, the system may
determine alternate deterministic paths for those flows that
predictably vary in terms of traffic volumes over time. By
rerouting a deterministic flow over one of these paths, physical
resources may be freed for a given duration for use by another
deterministic flow (e.g., a new flow, or a flow moved as part of a
same transaction). It is noted that rerouting over a longer path
consumes a higher amount of resources in the network, and should be
avoid under normal (non-peak) conditions.
The techniques herein also allow deterministic flows to be rerouted
during particular periods of time, so as to avoid choking
best-effort traffic. As a side effect, a slightly higher amount of
deterministic flows may also be admitted in the network, if flows
that do not fully use their committed rate are moved to make way at
particular times. Said differently, after identifying seasonal
flows, the supervisory device can move these seasonal flows to
links and nodes so that their low tides (e.g., when their traffic
volumes are at a minimum) match high volumes of best effort
traffic, and vice-versa. More stable flows are preferably placed in
links where the best effort traffic is less dynamic. In some
embodiments, this may be performed by a Predictive Time Schedule
PCE that positions the flows across the network and moves them
based on the predicted peak loads of best effort traffic.
As would be appreciated, the relocating of deterministic flows
along alternate deterministic paths can be performed within a
deterministic TSCH network by adjusting the timeslot assignments,
accordingly. In a similar manner, a switch (e.g. a TSN switch)
would be reconfigured to switch to a different port at potentially
a slightly different time. In further embodiments, this approach
can also be implemented in non-TSCH networks, and even wired
networks, by redirecting seasonal deterministic traffic along
different paths, to accommodate peak volumes of best effort
traffic.
FIG. 11 illustrates an example simplified procedure for
predictively adjusting time slot assignment in accordance with one
or more embodiments described herein. Procedure 1100 may be
implemented, for example, by a prediction engine/centralized
networking device, such as a PCE. Notably, the device may be
centralized in the sense that it may oversee the operation of other
network devices and may not be `centralized` from a geographical
standpoint. The procedure 1100 may start at step 1105, and
continues to step 1110, where, as described in greater detail
above, time slot usage reports may be received by the device. For
example, a time slot usage report may indicate which cells of a CDU
matrix (e.g., timeslots and associated channels) were used by a
network node. In some cases, a time slot usage report may be
received from a parent node in the network that monitors the usage
of time slots allocated to its child nodes. In one embodiment, a
time slot usage report may also be based on a notification sent by
a child node that indicates that the child node has experienced a
queuing delay.
At step 1115, as described in greater detail above, a time slot
demand change for a particular node is predicted based on the one
or more time slot usage reports. According to various embodiments,
the time slot usage reports received in step 1110 may be used as
input to a machine learning predictive model. Example models may
include, but are not limited to, ARMA models, ARMA-X models, HMMs,
Gaussian Processes, or any other machine learning model that can be
used to predict traffic demand changes and/or the seasonality of
such changes. For example, the device may predict that a particular
node will generate a spike in network traffic and corresponding
demand for TSCH time slots during a specific time of day. Such a
predicted demand change may also be predicted to be periodic, based
on previous time slot usage.
At step 1120, a time frame associated with the predicted time slot
demand change may be identified, as described in greater detail
above. For example, an increase or decrease in time slot demand by
a particular node may be predicted to begin at a specific point in
time and last for a predicted duration. In some cases, the time
frame may also be open ended. For example, a time frame may
indicate a start time associated with the change in time slot
demand, but not have a corresponding end time (e.g., the change is
predicted to be permanent).
At step 1125, one or more time slot assignments for the node(s)
predicted to experience a time slot demand change may be adjusted
based on the predicted time slot demand change and associated time
frame. For example, if a particular node is predicted to need more
time slot allocations than are currently allocated, it may be
allocated extra time slots either preemptively or at a time
associated with the predicted demand change. In one embodiment, the
time slot assignments may be made explicitly by the device to the
node(s). In another embodiment, the time slot adjustments may be
provided to a parent node of the node(s). For example, the
centralized device may notify the parent node of the predicted time
slot demand change and associated time period, thereby causing the
parent to generate an updated time slot assignment for the node(s)
to use during the time period. Procedure 1100 then ends at a step
1130. In some embodiments, procedure may be repeated any number of
times as part of a closed-loop mechanism whereby the central device
receives feedback regarding the time slot adjustment and makes
further adjustments as needed.
FIG. 12 illustrates an example simplified procedure for adjusting
time slot assignments of one or more child nodes in accordance with
one or more embodiments described herein. Procedure 1200 may be
implemented, for example, by a network device/node. The procedure
1200 may start at step 1205, and continues to step 1210, where, as
described in greater detail above, a time slot usage report may be
provided to a prediction engine (e.g., a centralized network
device). The usage report may be generated, for example, by
monitoring the use of TSCH time slots assigned to the one or more
child nodes of the device. In one embodiment, the time slot usage
reports may also be based on notifications received by the child
node(s) regarding any queuing delays experienced by a child
node.
At step 1215, as described in greater detail above, a predicted
time slot usage change for the one or more child nodes is received
from the prediction engine. The change may indicate, for example,
that a given child node is predicted to need more or less time
slots (e.g., an amount of time slots) during a specified time
period. For example, the received change may indicate that the
child node is predicted to need additional time slots beginning at
a specific point in time.
At step 1220, updated time slot assignments(s) are generated for
the one or more child nodes, as detailed above, based on the
predicted time slot usage change. For example, based on a
prediction that a child node will need additional time slots
starting at a certain point in time, the parent node may increase
the number of time slot assignments to the child node either at the
point in time or before the point in time (e.g., proactively). In
one embodiment, arbitration may be performed among child nodes such
that time slots are reassigned from one node predicted to have
fewer demands to a child node predicted to experience an influx of
traffic.
At step 1225, the updated time slot assignment(s) are provided to
the child node(s), as described in greater detail above. For
example, the parent node that owns a chunk of the overall TSCH
schedule may reallocate time slot assignments to its child node(s).
In one embodiment, the assignment(s) may be provided at a time that
also takes into account the delay associated with the reassignment.
For example, if a time slot demand change is predicted to occur at
a time T2 for a child node, the corresponding update to the time
slot assignments of the child node may be initiated prior to time
T2, to account for the reassignment process. Procedure 1200 then
ends at step 1230.
FIG. 13 illustrates an example simplified procedure for generating
a time slot usage report in accordance with one or more embodiments
described herein. Procedure 1300 may be implemented, for example,
by a parent network device/node of one or more child node(s) in the
network. The procedure 1300 may start at step 1305, and continues
to step 1310, where, as described in greater detail above, the
parent node may monitor time slot usage by its one or more child
nodes. For example, the parent node may determine whether or not a
child node uses an assigned TSCH time slot to communicate with the
parent node.
At step 1315, delay notification(s) are received from the one or
more child nodes, as described in greater detail above. As noted
previously, a parent node may not be able to determine whether a
child node is experiencing queuing delays by simply observing the
time slots used by the child node. In some embodiments, the child
node may send a notification to the parent node that indicates that
the child node has delayed sending some traffic until another time
slot. The notification may also indicate a traffic priority for the
queued traffic, which may be used as part of the decision to adjust
the time slots allocated to the child node. For example, queuing
delays associated with high priority traffic may be a greater
indicator that more time slots should be allocated to the child
node, whereas queuing delays with low priority traffic may be more
acceptable.
At step 1320, a time slot usage report is generated, as described
in greater detail above. As noted above, such a report may include
information regarding the time slot usage by the child node(s), as
well as any delays reported by the child node(s). For example, such
a report may indicate that a particular child node is using all of
its allocated time slots, but is still experiencing queuing delays,
thereby indicating that the child node may need additional time
slots.
At step 1325, as detailed above, the generated time slot usage
report may be provided to a predictive time scheduler. For example,
the usage report may be provided to a PTS executed by a centralized
networking device, such as a PCE, NMS, etc. In response, the PTS
may use the reports to predict future time slot demands for the
network nodes and proactively initiate changes to their time slot
assignments. Procedure 1300 then ends at step 1330.
FIG. 14 illustrates an example simplified procedure for moving
seasonal deterministic traffic between network paths, in accordance
with one or more embodiments described herein. Procedure 1400 may
be implemented, for example, by a prediction engine/centralized
networking device, such as a PCE, or another form of supervisory
network devices. Procedure 1400 may start at step 1405, and
continues on to step 1410 where, as described in greater detail
above, the device may receive data regarding traffic volumes of
deterministic and non-deterministic traffic along a first path in
the network. Such data may comprise, for example, Netflow traffic
records, IPFIX traffic records, or the like.
At step 1415, as detailed above, the device may predict, using the
received data, an increase in the traffic volume of the
non-deterministic traffic along the first path in the network. In
some embodiments, the device may do so using a machine learning
process, such as an ARMA model, ARMA-X model, HMM, Gaussian
Process, or any other machine learning model that can be used to
predict seasonal/periodic volume changes in the traffic along the
first path.
At step 1420, the device may identify a period of time associated
with the predicted increase in the traffic volume of the
non-deterministic traffic along the first path, as described in
greater detail above. For example, the non-deterministic traffic
may be predicted to increase during a certain time of day, on a
certain day, or the like.
At step 1425, as detailed above, the device may cause the
deterministic traffic to be sent along a second path in the network
during the identified period of time, to allow the first path to
accommodate the predicted increase in the traffic volume of the
non-deterministic traffic along the first path. In various
embodiments, the device may do so, in part, by unreserving any
resources reserved for the deterministic traffic during the
identified period of time. Such resource may include, for example,
reserved bandwidth, communication time slots, or the like. After
the time period is over, the device may further cause the
deterministic traffic to return to the first path, in some cases.
Procedure 1400 then ends at step 1430.
It should be noted that while certain steps within procedures
1100-1400 may be optional as described above, the steps shown in
FIGS. 11-14 are merely examples for illustration, and certain other
steps may be included or excluded as desired. Further, while a
particular order of the steps is shown, this ordering is merely
illustrative, and any suitable arrangement of the steps may be
utilized without departing from the scope of the embodiments
herein. Moreover, while procedures 1100-1400 are described
separately, certain steps from each procedure may be incorporated
into each other procedure, and the procedures are not meant to be
mutually exclusive.
The techniques described herein, therefore, provide for an
architecture that may dramatically increase the scalability of
time-based scheduling approaches in TSCH networks. The techniques
herein may also improve the quality of service (QoS) of real-time
flows since predicted traffic increases or decrease, and
potentially the seasonality of such changes, may be used to
proactively adjust the time slot allocations of the network nodes.
Furthermore, since the time slot allocations can be proactively
initiated, the heavy computations associated with the predictions
may be performed in the background instead of being triggered
reactively (e.g., based on real-time reporting of network
conditions). In addition, the techniques herein may considerably
reduce the overhead of the control plane, which may be of utmost
importance in constrained networks, such as LLNs implementing TSCH.
While there have been shown and described illustrative embodiments
that provide for the arbitration of time contention in a
shared-media communication network, it is to be understood that
various other adaptations and modifications may be made within the
spirit and scope of the embodiments herein. For example, the
embodiments have been shown and described herein primarily with
respect to LLNs. However, the embodiments in their broader sense
are not as limited, and may, in fact, be used with other types of
networks and/or protocols (e.g., wireless). In addition, while
certain protocols are shown, such as RPL, other suitable protocols
may be used, accordingly.
The foregoing description has been directed to specific
embodiments. It will be apparent, however, that other variations
and modifications may be made to the described embodiments, with
the attainment of some or all of their advantages. For instance, it
is expressly contemplated that the components and/or elements
described herein can be implemented as software being stored on a
tangible (non-transitory) computer-readable medium (e.g.,
disks/CDs/RAM/EEPROM/etc.) having program instructions executing on
a computer, hardware, firmware, or a combination thereof.
Accordingly this description is to be taken only by way of example
and not to otherwise limit the scope of the embodiments herein.
Therefore, it is the object of the appended claims to cover all
such variations and modifications as come within the true spirit
and scope of the embodiments herein.
* * * * *